X-no-Archive:yes
To answer this, I think we have to first set aside how we feel about hp's
decision, and try to think, as objectively as we can, about what the market
is doing, which is a very different thing from what we work with every day.
If I may offer an analogy, this is perhaps like the difference between the
car you drive, and what the dealers are selling today or promising tomorrow,
for those of us who are not driving this year's model.

Wirt wrote about blade servers, at one end of the market, where the product
is a commodity. I recently wrote that hp seems to betting that the industry
is moving toward commoditization, where we will buy IT products and services
the way we buy groceries and gasoline. If Wirt's right about hp's strategy,
and hp gets there first, then they are the brand of frosted flakes with the
tiger on the box. See for yourself at
http://www.hp.com/products1/servers/blades/index.html. And note that this
page mentions "the Itanium processor family running Linux, HP-UX, and
Windows".

Christian wrote about products which we can buy from hp, some part of which
are bought from someone else. But I think that most if not all of us have
had this experience, arguing over whose product is causing the problem.
Here, the benefits of single source become obvious: you sold it to me, you
fix the problem and make it work, within some contractual SLA (and this is
one area where the need for improved professional services comes in).

Better than this full support from a single supplier, and back to what Wirt
wrote about blades: "If a unit fails, you simply pull it out and replace it
with a twin spare unit, potentially made by anyone, but preferentially by
HP." Furthermore, if these are redundant, clustered or load balanced for
instance, and you lose one of five, then you are running at 80%, instead of
0%, with a pretty good chance of no one else being able to much tell the
difference. And if hp sold you the blades and everything with it, you don't
even have to figure out exactly what failed. You just call hprc.

And, in n months, when the new version or model comes out, will it continue
to work with last year's model? If you've bought exclusively from hp, then
hp has a vested interest in making sure that the new model integrates as
seamlessly as possible with what you already have. This is even  more
important for the unit failure scenario. And it relieves any burden to stock
replacement parts - replacement parts are the new version, and if it works,
who's going to complain about getting a newer, faster, better product to
replace the failed unit?

But at some point, products and systems mature, and the number of supported
systems may grow to a point where management begins to look at
consolidation. Instead of backing up four dozen blades, does it make sense
to have fewer, large, dedicated servers? Is there a point at which something
like SuperDome begins to make more sense than a rack farm? IBM is touting
this very approach in a commercial (production values aside). While hp
wasn't there the firstest with the mostest, if they sold you all those
blades, they'll probably sell you anything you want.

Now, that Winston Prather and others from CSY have already moved over to
this effort, that seems to be a tale of intrigue.

Greg Stigers
http://www.cgiusa.com
Share and Enjoy - [log in to unmask]

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *